DeepMind puts the entire human proteome online, as folded by AlphaFold

DeepMind and several research partners have released a database containing the 3D structures of nearly every protein in the human body, as computationally determined by the breakthrough protein folding system demonstrated last year, AlphaFold. The freely available database represents an enormous advance and convenience for scientists across hundreds of disciplines and domains, and may very well form the foundation of a new phase in biology and medicine.

The AlphaFold Protein Structure Database is a collaboration between DeepMind, the European Bioinformatics Institute and others, and consists of hundreds of thousands of protein sequences with their structures predicted by AlphaFold — and the plan is to add millions more to create a “protein almanac of the world.”

“We believe that this work represents the most significant contribution AI has made to advancing the state of scientific knowledge to date, and is a great example of the kind of benefits AI can bring to society,” said DeepMind founder and CEO Demis Hassabis.

From genome to proteome

If you’re not familiar with proteomics in general — and it’s quite natural if that’s the case — the best way to think about this is perhaps in terms of another major effort: that of sequencing the human genome. As you may recall from the late ’90s and early ’00s, this was a huge endeavor undertaken by a large group of scientists and organizations across the globe and over many years. The genome, finished at last, has been instrumental to the diagnosis and understanding of countless conditions, and in the development of drugs and treatments for them.

It was, however, just the beginning of the work in that field — like finishing all the edge pieces of a giant puzzle. And one of the next big projects everyone turned their eyes toward in those years was understanding the human proteome — which is to say all the proteins used by the human body and encoded into the genome.

The problem with the proteome is that it’s much, much more complex. Proteins, like DNA, are sequences of known molecules; in DNA these are the handful of familiar bases (adenine, guanine, etc.), but in proteins they are the 20 amino acids (each of which is coded by multiple bases in genes). This in itself creates a great deal more complexity, but it’s only the start. The sequences aren’t simply “code” but actually twist and fold into tiny molecular origami machines that accomplish all kinds of tasks within our body. It’s like going from binary code to a complex language that manifests objects in the real world.

Practically speaking this means that the proteome is made up of not just 20,000 sequences of hundreds of acids each, but that each one of those sequences has a physical structure and function. And one of the hardest parts of understanding them is figuring out what shape is made from a given sequence. This is generally done experimentally using something like x-ray crystallography, a long, complex process that may take months or longer to figure out a single protein — if you happen to have the best labs and techniques at your disposal. The structure can also be predicted computationally, though the process has never been good enough to actually rely on — until AlphaFold came along.

Taking a discipline by surprise

Without going into the whole history of computational proteomics (as much as I’d like to), we essentially went from distributed brute-force tactics 15 years ago — remember [email protected]? — to more honed processes in the last decade. Then AI-based approaches came on the scene, making a splash in 2019 when DeepMind’s AlphaFold leapfrogged every other system in the world — then made another jump in 2020, achieving accuracy levels high enough and reliable enough that it prompted some experts to declare the problem of turning an arbitrary sequence into a 3D structure solved.

I’m only compressing this long history into one paragraph because it was extensively covered at the time, but it’s hard to overstate how sudden and complete this advance was. This was a problem that stumped the best minds in the world for decades, and it went from “we maybe have an approach that kind of works, but extremely slowly and at great cost” to “accurate, reliable, and can be done with off the shelf computers” in the space of a year.

Examples of protein structures predicted by AlphaFold

Image Credits: DeepMind

The specifics of DeepMind’s advances and how it achieved them I will leave to specialists in the fields of computational biology and proteomics, who will no doubt be picking apart and iterating on this work over the coming months and years. It’s the practical results that concern us today, as the company employed its time since the publication of AlphaFold 2 (the version shown in 2020) not just tweaking the model, but running it… on every single protein sequence they could get their hands on.

The result is that 98.5% of the human proteome is now “folded,” as they say, meaning there is a predicted structure that the AI model is confident enough (and importantly, we are confident enough in its confidence) represents the real thing. Oh, and they also folded the proteome for 20 other organisms, like yeast and E. coli, amounting to about 350,000 protein structures total. It’s by far — by orders of magnitude — the largest and best collection of this absolutely crucial information.

All that will be made available as a freely browsable database that any researcher can simply plug a sequence or protein name into and immediately be provided the 3D structure. The details of the process and database can be found in a paper published today in the journal Nature.

“The database as you’ll see it tomorrow, it’s a search bar, it’s almost like Google search for protein structures,” said Hassabis in an interview with TechCrunch. “You can view it in the 3D visualizer, zoom around it, interrogate the genetic sequence… and the nice thing about doing it with EMBL-EBI is it’s linked to all their other databases. So you can immediately go and see related genes, And it’s linked to all these other databases, you can see related genes, related in other organisms, other proteins that have related functions, and so on.”

Image Credits: DeepMind

“As a scientist myself, who works on an almost unfathomable protein,” said EMBL-EBI’s Edith Heard (she didn’t specify which protein), “it’s really exciting to know that you can find out what the business end of a protein is now, in such a short time — it would have taken years. So being able to access the structure and say ‘aha, this is the business end,’ you can then focus on trying to work out what that business end does. And I think this is accelerating science by steps of years, a bit like being able to sequence genomes did decades ago.”

So new is the very idea of being able to do this that Hassabis said he fully expects the entire field to change — and change the database along with it.

“Structural biologists are not yet used to the idea that they can just look up anything in a matter of seconds, rather than take years to experimentally determine these things,” he said. “And I think that should lead to whole new types of approaches to questions that can be asked and experiments that can be done. Once we start getting wind of that, we may start building other tools that cater to this sort of serendipity: What if I want to look at 10,000 proteins related in a particular way? There isn’t really a normal way of doing that, because that isn’t really a normal question anyone would ask currently. So I imagine we’ll have to start producing new tools, and there’ll be demand for that once we start seeing how people interact with this.”

That includes derivative and incrementally improved versions of the software itself, which has been released in open source along with a great deal of development history. Already we have seen an independently developed system, RoseTTAFold, from researchers at the University of Washington’s Baker Lab, which extrapolated from AlphaFold’s performance last year to create something similar yet more efficient — though DeepMind seems to have taken the lead again with its latest version. But the point was made that the secret sauce is out there for all to use.

Practical magic

Although the prospect of structural bioinformaticians attaining their fondest dreams is heartwarming, it is important to note that there are in fact immediate and real benefits to the work DeepMind and EMBL-EBI have done. It is perhaps easiest to see in their partnership with the Drugs for Neglected Diseases Institute.

The DNDI focuses, as you might guess, on diseases that are rare enough that they don’t warrant the kind of attention and investment from major pharmaceutical companies and medical research outfits that would potentially result in discovering a treatment.

“This is a very practical problem in clinical genetics, where you have a suspected series of mutations, of changes in an affected child, and you want to try and work out which one is likely to be the reason why our child has got a particular genetic disease. And having widespread structural information, I am almost certain will improve the way we can do that,” said DNDI’s Ewan Birney in a press call ahead of the release.

Ordinarily examining the proteins suspected of being at the root of a given problem would be expensive and time-consuming, and for diseases that affect relatively few people, money and time are in short supply when they can be applied to more common problems like cancers or dementia-related diseases. But being able to simply call up the structures of 10 healthy proteins and 10 mutated versions of the same, insights may appear in seconds that might otherwise have taken years of painstaking experimental work. (The drug discovery and testing process still takes years, but maybe now it can start tomorrow for Chagas disease instead of in 2025.)

Illustration of RNA polymerase II ( a protein) in action in yeast. Image Credits: Getty Images / JUAN GAERTNER/SCIENCE PHOTO LIBRARY

Lest you think too much is resting on a computer’s prediction of experimentally unverified results, in another, totally different case, some of the painstaking work had already been done. John McGeehan of the University of Portsmouth, with whom DeepMind partnered for another potential use case, explained how this affected his team’s work on plastic decomposition.

“When we first sent our seven sequences to the DeepMind team, for two of those we already had experimental structures. So we were able to test those when they came back, and it was one of those moments, to be honest, when the hairs stood up on the back of my neck,” said McGeehan. “Because the structures that they produced were identical to our crystal structures. In fact, they contained even more information than the crystal structures were able to provide in certain cases. We were able to use that information directly to develop faster enzymes for breaking down plastics. And those experiments are already underway, immediately. So the acceleration to our project here is, I would say, multiple years.”

The plan is to, over the next year or two, make predictions for every single known and sequenced protein — somewhere in the neighborhood of a hundred million. And for the most part (the few structures not susceptible to this approach seem to make themselves known quickly) biologists should be able to have great confidence in the results.

Inspecting molecular structure in 3D has been possible for decades, but finding that structure in the first place is difficult. Image Credits: DeepMind

The process AlphaFold uses to predict structures is, in some cases, better than experimental options. And although there is an amount of uncertainty in how any AI model achieves its results, Hassabis was clear that this is not just a black box.

“For this particular case, I think explainability was not just a nice-to-have, which often is the case in machine learning, but it was a must-have, given the seriousness of what we wanted it to be used for,” he said. “So I think we’ve done the most we’ve ever done on a particular system to make the case with explainability. So there’s both explainability on a granular level on the algorithm, and then explainability in terms of the outputs, as well the predictions and the structures, and how much you should or shouldn’t trust them, and which of the regions are the reliable areas of prediction.”

Nevertheless, his description of the system as “miraculous” attracted my special sense for potential headline words. Hassabis said that there’s nothing miraculous about the process itself, but rather that he’s a bit amazed that all their work has produced something so powerful.

“This was by far the hardest project we’ve ever done,” he said. “And, you know, even when we know every detail of how the code works, and the system works, and we can see all the outputs, it’s still just still a bit miraculous when you see what it’s doing… that it’s taking this, this 1D amino acid chain and creating these beautiful 3D structures, a lot of them aesthetically incredibly beautiful, as well as scientifically and functionally valuable. So it was more a statement of a sort of wonder.”

Fold after fold

The impact of AlphaFold and the proteome database won’t be felt for some time at large, but it will almost certainly — as early partners have testified — lead to some serious short-term and long-term breakthroughs. But that doesn’t mean that the mystery of the proteome is solved completely. Not by a long shot.

As noted above, the complexity of the genome is nothing compared to that of the proteome at a fundamental level, but even with this major advance we have only scratched the surface of the latter. AlphaFold solves a very specific, though very important problem: given a sequence of amino acids, predict the 3D shape that sequence takes in reality. But proteins don’t exist in a vacuum; they’re part of a complex, dynamic system in which they are changing their conformation, being broken up and reformed, responding to conditions, the presence of elements or other proteins, and indeed then reshaping themselves around those.

In fact a great deal of the human proteins for which AlphaFold gave only a middling level of confidence to its predictions may be fundamentally “disordered” proteins that are too variable to pin down the way a more static one can be (in which case the prediction would be validated as a highly accurate predictor for that type of protein). So the team has its work cut out for it.

“It’s time to start looking at new problems,” said Hassabis. “Of course, there are many, many new challenges. But the ones you mentioned, protein interaction, protein complexes, ligand binding, we’re working actually on all these things, and we have early, early stage projects on all those topics. But I do think it’s worth taking, you know, a moment to just talk about delivering this big step… it’s something that the computational biology community’s been working on for 20, 30 years, and I do think we have now broken the back of that problem.”

FTC puts hardware makers on warning for potential ‘unlawful repair restrictions’

As phones and other consumer devices have gained feature after feature, they have also declined in how easily they can be repaired, with Apple at the head of this ignoble pack. The FTC has taken note, admitting that the agency has been lax on this front but that going forward it will prioritize what could be illegal restrictions by companies as to how consumers can repair, repurpose, and reuse their own property.

Devices are often built today with no concessions made towards easy repair or refurbishment, or even once routine upgrades like adding RAM or swapping out an ailing battery. While companies like Apple do often support hardware for a long time in some respects, the trade-off seems to be that if you crack your screen, the maker is your only real option to fix it.

That’s a problem for many reasons, as right-to-repair activist and iFixit founder Kyle Wiens has argued indefatigably for years (the company posted proudly about the statement on its blog). The FTC sought comment on this topic back in 2019, issued a report on the state of things a few months ago, and now (perhaps emboldened by new Chair Lina Khan’s green light to all things fearful to big tech companies) has issued a policy statement.

The gist of the unanimously approved statement is that they found that the practice of deliberately restricting repairs may have serious repercussions, especially among people who don’t have the cash to pay the Apple tax for what ought to be (and once was) a simple repair.

The Commission’s report on repair restrictions explores and discusses a number of these issues and describes the hardships repair restrictions create for families and businesses. The Commission is concerned that this burden is borne more heavily by underserved communities, including communities of color and lower-income Americans. The pandemic exacerbated these effects as consumers relied more heavily on technology than ever before.

While unlawful repair restrictions have generally not been an enforcement priority for the Commission for a number of years, the Commission has determined that it will devote more enforcement resources to combat these practices. Accordingly, the Commission will now prioritize investigations into unlawful repair restrictions under relevant statutes…

The statement then makes four basic points. First, it reiterates the need for consumers and other public organizations to report and characterize what they perceive as unfair or problematic repair restrictions. The FTC doesn’t go out and spontaneously investigate companies, it generally needs a complaint to set the wheels in motion, such as people alleging that Facebook is misusing their data.

Second is a surprising antitrust tie-in, where the FTC says it will look at said restrictions aiming to answer whether monopolistic practices like tying and exclusionary design are in play. This could be something like refusing to allow upgrades, then charging an order of magnitude higher than market price for something like a few extra gigs of storage or RAM, or designing products in such a way that it moots competition. Or perhaps arbitrary warranty violations for doing things like removing screws or taking the device to third party for repairs. (Of course, these would depend on establishing monopoly status or market power for the company, something the FTC has had trouble doing.)

More in line with the FTC’s usual commercial regulations, it will assess whether the restrictions are “unfair acts or practices,” which is a much broader and easier to meet requirement. You don’t need a monopoly to make claims of an “open standard” to be misleading, or for a hidden setting to slow the operations of third party apps or peripherals, for instance.

And lastly the agency mentions that it will be working with states in its push to establish new regulations and laws. This is perhaps a reference to the pioneering “right to repair” bills like the one passed by Massachusetts last year. Successes and failures along those lines will be taken into account and the feds and state policymakers will be comparing notes.

This isn’t the first movement in this direction by a long shot, but it is one of the plainest. Tech companies have seen the writing on the wall, and done things like expand independent repair programs — but it’s arguable that these actions were taken in anticipation of the FTC’s expected shift toward establishing hard lines on the topic.

The FTC isn’t showing its full hand here, but it’s certainly hinting that it’s ready to play if the companies involved want to push their luck. We’ll probably know more soon once it starts ingesting consumer complaints and builds a picture of the repair landscape.

Pivot Bio rakes in $430M round D as modified microbes prove their worth in agriculture

Pivot Bio makes fertilizer — but not directly. Its modified microorganisms are added to soil and they product nitrogen that would otherwise have had to be trucked in and dumped there. This biotech-powered approach can save farmers money and time and ultimately may be easier on the environment — a huge opportunity that investors have plowed $430 million into in the company’s latest funding round.

Nitrogen is among the nutrients crops need to survive and thrive, and it’s only by dumping fertilizer on the soil and mixing it in that farmers can keep growing at today’s rates. But in some ways we’re still doing what our forebears did generations ago.

“Fertilizer changed agriculture — it’s what made so much of the last century possible. But it’s not a perfect way to get nutrients to crops,” said Karsten Temme, CEO and co-founder of Pivot Bio. He pointed out the simple fact that distributing fertilizer over a thousand — let alone ten thousand or more — acres of farmland is an immense mechanical and logistical challenge, involving many people, heavy machinery, and valuable time.

Not to mention the risk that a heavy rain might carry off a lot of the fertilizer before it’s absorbed and used, and the huge contributions of greenhouse gases the fertilizing process produces. (The microbe approach seems to be considerably better for the environment.)

Yet the reason we do this in the first place is essentially to imitate the work of microbes that live in the soil and produce nitrogen naturally. Plants and these microbes have a relationship going back millions of years, but the tiny organisms simply don’t produce enough. Pivot Bio’s insight when it started more than a decade ago was that a few tweaks could supercharge this natural nitrogen cycle.

“We’ve all known microbes were the way to go,” he said. “They’re naturally part of the root system — they were already there. They have this feedback loop, where if they detect fertilizer they don’t make nitrogen, to save energy. The only thing that we’ve done is, the portion of their genome responsible for producing nitrogen is offline, and we’re waking it up.”

Other agriculture-focused biotech companies like Indigo and AgBiome are also looking at modifying and managing the plant’s “microbiome,” which is to say the life that lives in the immediate vicinity of a given plant. A modified microbiome may be resistant to pests, reduce disease, or offer other benefits.

Illustration showing stages of modifying and deploying nitrogen-producing microbes.

Image Credits: Pivot Bio

It’s not so different from yeast, which as many know from experience works as a living rising agent. That microbe has been cultivated to consume sugar and produce a gas, which leads to the air pockets in baked goods. This microbe has been modified a bit more directly to continually consume the sugars put out by plants and put out nitrogen. And they can do it at rates that massively reduce the need for adding solid fertilizer to the soil.

“We’ve taken what is traditionally tons and tons of physical materials, and shrunk that into a powder, like baker’s yeast, that you can fit in your hand,” Temme said (though, to be precise, the product is applied as a liquid). “All of a sudden managing that farm gets a little easier. You free up the time you would have spent sitting in the tractor applying fertilizer to the field; you’ll add our product at the same time you’d be planting your seeds. And you have the confidence that if a rainstorm comes through in the spring, it’s not washing it all away. Globally about half of all fertilizer is washed away… but microbes don’t mind.”

Instead, the microbes just quietly sit in the soil pumping out nitrogen at a rate of up to 40 pounds per acre — a remarkably old-fashioned way to measure it (why not grams per square centimeter?) but perhaps in keeping with agriculture’s occasional anachronistic tendencies. Depending on the crop and environment that may be enough to do without added fertilizers at all, or it might be about half or less.

Whatever the proportion provided by the microbes, it must be tempting to employ them, because Pivot Bio tripled its revenue in 2021. You might wonder why they can be so sure only halfway through the year, but as they are currently only selling to farmers in the northern hemisphere and the product is applied at planting time early in the year, they’re done with sales for the year and can be sure it’s three times what they sold in 2020.

The microbes die off once the crop is harvested, so it’s not a permanent change to the ecosystem. And next year, when farmers come back for more, the organisms may well have been modified further. It’s not quite as simple as turning the nitrogen production on or off in the genome; the enzymatic pathway from sugar to nitrogen can be improved, and the threshold for when the microbes decide to undertake the process rather than rest can be changed as well. The latest iteration, Proven 40, has the yield mentioned above, but further improvements are planned, attracting potential customers on the fence about whether it’s worth the trouble to change tactics.

The potential for recurring revenue and growth (by their current estimate they are currently able to address about a quarter of a $200 billion total market) led to the current monster D round, which was led by DCVC and Temasek. There are about a dozen other investors, for which I refer readers to the press release, which lists them in no doubt a very carefully negotiated order.

Temme says the money will go towards deepening and broadening the platform and growing the relationship with farmers, who seem to be hooked after giving it a shot. Right now the microbes are specific to corn, wheat, and rice, which of course covers a great deal of agriculture, but there are many other corners of the industry that would benefit from a streamlined, enhanced nitrogen cycle. And it’s certainly a powerful validation of the vision Temme and his co-founder Alvin Tamsir had 15 years ago in grad school, he said. Here’s hoping that’s food for thought for those in that position now, wondering if it’s all worth it.

Virgin Galactic president Mike Moses on what’s next for the company’s growing fleet

This last weekend featured the much-ballyhooed launch of Virgin Galactic’s first (nonpaying) passengers, with founder and CEO Richard Branson along for the ride. After the festivities, I had the chance to talk with the company’s president, Mike Moses, who seems to be familiar with every detail of the operation and the company’s plans for going from test to commercial flights.

Unfortunately my recorder went on the fritz, but Moses was kind enough to hop on the phone later in the week to talk (again) about the next generation of spaceplanes, where the company needs to invest, and more. You can read through our conversation below. (Interview has been lightly edited for clarity.)

TC: To begin with, can you tell me what’s left to test, and when do you expect to finish the test flight phase?

Moses: The test flight series that we’re kind of in right now, and the flight with Richard was the first of those, represents a shift from what was more classic and traditional, envelope testing, where we’re looking at aerodynamics and trajectories and handling qualities, to more of an operational check-out, where we are validating cabin experience experiences, training procedures, hardware for the folks in the back and what they’re going to go through.

So we’ve laid out a series of a few flights there, three to be specific, that both demonstrate key product milestones and features, as well as allow us time to iterate and develop and optimize some of that back-of-cabin experience. But as always, that’s a notional schedule, right? The schedule and the numbers are going to depend on the results. So if things go well, we think that’s a three-flight series if we find things that we need to adjust, we’ll add more as needed based on what we’re learning.

Based on the results that we got after Richard and crew came back from the last flight, you know, we know we have some stuff to work on but but everything was pretty much thumbs up.

Now, we know we’re going to do those flights over the course of this summer and late summer, and then we’ll be ready to move into, as we announced during our previous earnings call, a ‘modification phase’ where we’re going to do some upgrades on our mothership and our spaceship to prepare them for commercial service. The main focus there is to look at things that allow us to increase the flight-rate frequency. Right now in test, we fly at a fairly slow pace [i.e. infrequently, not at low speed], because we’re inspecting everything prudently. We’re going to want to start to move away from that, and as we learn, and so we already know, there’s some modifications we want to make to enable that to start to happen. We haven’t set a specific timescale for when that officially ends.

TC: You mentioned when we talked at the Spaceport, the crew hadn’t yet really been debriefed about the experience. I’m hoping maybe you have a little more information now about recommendations from Sir Richard, from Siriha, from everybody that was actually up there. Have you gotten any substantive feedback that you can share?

Moses: So we are definitely in the middle of all that feedback and debriefing. As you might imagine, there’s a lot of data to go through. And in some cases, that data is as simple as the 16 video cameras that we had onboard, and getting them all synced up to see that what’s happening where, and couple that with live notes, and debriefs, and the audio tracks that went with it. We are definitely gathering up the inputs, but there’s nothing on that list that I think I’m ready to disclose at this time. We’ll keep folks posted as we go.

The general feedback, post-landing both that day and the next day, was ‘things were awesome,’ right? Now that’s not a scientific answer, and I want the scientific answer, so we’re gonna make them go through the work to debrief.

Image Credits: Virgin Galactic

TC: You touched on this with the ‘modification phase’… Unity is, I don’t know how exactly you’d describe it, a production prototype. Could you tell me whether there’s any special upkeep for it as the sort of first off the line?

Moses: There’s nothing special as part of its design or build that requires special upkeep. But as a test vehicle and as our first article, we give it a lot of extra attention. We dive in pretty deep on inspections, both regularly and as we see issues, we would probably, test those and explore just to make sure we truly understand that there’s no unknowns out there, things like how the system performs how it does in cold temperatures, under load and under stress. We keep an eye on it.

There’s a series of measurements that we make to say, you know, where did the vehicle perform based on its design envelope. And if we’re close to the edges of any of that envelope, we go do extra inspections to validate that our modeling and our predictions are right. So in that regard, it’s pretty similar to how you would have a first set of articles coming out for a new aircraft development, you would build a maintenance and inspection program. That is, an extremely conservative one. And then as you use it, you start to pull out that conservatism based on your positive feedback.

But in general, yes, Unity does get a lot of extra attention. And the next vehicles will have some of that designed in part of that. We’ve already learned a bunch of, like, ‘hey, on the next vehicle, make this different so I don’t have to look at it every time, I can look at it every five times.’

TC: I think that when we when we talked before, you mentioned that you expect multiple-hundred flights, at least theoretically, out of Unity.

Moses: Yeah, multiple-hundred flights of the vehicle. We set a design envelope where we designed for a certain lifetime, and we we tested to that, and then we can always go do life extension. Some of that is just a limitation of… you know, we’re going to cycle the stuff 10,000 times rather than 40,000 times, and we’ll come back later and get the other cycles when we get closer to the 10,000 life. We’ll go back and add more to it. There’s not a lot of components that have, you know, like a ‘fall off the cliff’ type of lifetime.

TC: You mentioned some of the modifications you are going to build into the successor or production craft. Can you tell me any of those, how it will differ in minor or major ways, when you expect weight on wheels and that kind of thing?

Moses: So we’ve already done weight on wheels. And we had our rollout, which is effectively that weight on wheels, where we transition from, basically major factory assembly into ground tests. So all of the systems are installed, and now they’re gonna start to run integrated ground testing, where you can basically go run a computer system through its checkouts, you can run the flight control system through checkouts… you’re still on the ground, right, you’re not yet ready to fly. But we are in that integrated testing.

As far as changes… when we designed the structure, if you think about it as the skeleton, under the skin, with Imagine and Inspire, we optimized and moved those skeletons, the ribs in the spars, to the locations where the load was highest. Unity was built off of the original design intent of Scaled Composites, and flight tests, they’ve shown us that sometimes that load is not exactly where it is expected. There’s a lot of extra weight in Unity to account for that load; Imagine and Inspire, we’re able to optimize and put the structure right where it needed to be.

There’s a joint, for example, on Unity that I have to go look at every time, because I had to add extra to it. Whereas on Imagine, it was designed to where it should be in the first place. I’ll still look at it, but it’s much easier access and a much shorter inspection.

VSS Imagine on a runway.

Image Credits: Virgin Galactic

So things like that, that let me optimize my inspection schedule. And other just simplistic things — there are now access panels where we know we need them, whereas we had to kind of add them after the fact in Unity. Your quick release fasteners and things like that, that make inspections shorter, we were able to add into the design, we made a pretty significant number of changes like that, all fairly minor, but they have a large effect on the maintainability of the vehicle.

And the next phase, right, we talked about this, the Delta class of spaceships, we’re going to make changes for manufacturability. Unity and Inspire and Imagine are still fairly one-off hand-built aircraft — spacecraft, sorry. And if we want to go build a dozen or more to get to these 400-flight-a-year rates, we need to make sure they’re manufacturable at a smaller price tag in a smaller time scale. So that next design will incorporate a bunch of that stuff.

TC: That’s actually one of the things I wanted to talk about is how you get to the reliability and cadence that you want to have for commercial operation? Obviously, more aircraft is one part of that, but you know, maybe expanding ground ops or crew, better maintenance and stuff like that.

Moses: Yeah, you bet. And I think that’s it, right: It’s a fleet, so we have multiple vehicles for dispatch. That gives you capacity to be able to handle anything that comes up unexpected, like weather. And then it’s the workforce — with more workforce, a 24/7 clock, then you can have multiple expertises, or a crew focused on just one vehicle. And the second crew, they’re focused on the second one.

I think our mantra here is going to be to take it in baby steps — we’re not going to try to go to those high flight rates initially, we want to get a little faster, then a little faster, then a little faster. That’s kind of Unity’s purpose in life in 2022, to allow us to go explore those operational cadences and see where we can apply multiplying factors for when we get additional spaceships.

You know, the business model is a great one, right? But in these next couple of years, it’s fairly insensitive to whether I’m doing eight flights or 10 flights or 12 flights with Unity. I mean, in terms of revenue, it doesn’t move the needle very much. But in terms of operational learning, that’s a significant step for us, so we want to be prudent with how we proceed down that path.

MOJAVE, UNITED STATES – OCTOBER 10: (EDITORIAL USE ONLY, NO SUBJECT SPECIFIC TV BROADCAST DOCUMENTARIES OR BOOK USE) Virgin Galactic vehicle SpaceShipTwo completes its successful first glide flight at Mojave on October 10, 2010 over Mojave in California. (Photo by Mark Greenberg/Virgin Galactic/Getty Images)

TC: Can you can you tell me again why, or whether, you plan on keeping the flight plans more or less the same? Maybe there’s possibility, later down the line with the revised version with six people in it, that you might have to have a slightly different profile?

Moses: That’s kind of coupled with what we talked about at the beginning of this Q&A, the move from a test phase into this operational readiness phase. Coupled with that is a profile that is now set — the trajectory that the pilots fly, the techniques they use, we’ll still optimize, but we’re not making major revisions. Those are all pretty much physics-based results. The airspeed we’re at, the angles that we’re at, and the subsequent altitude we get to, the weight we carry, are all kind of locked-in variables, and there’s not much you can do to change that equation.

There’ll be some definite trajectory changes that come along with Imagine because it will have more capacity on board, which means it’ll have a slightly different performance, and we just need to go verify that envelope. But for the most part, you know, the physics of the equation kind of set what you can do, roughly speaking, so that’s why we’re limited to only carrying four passengers here initially. We can change that, and we do plan on looking at weight reductions in the ship, but again, with an eye towards the fleet that we’re building, and make sure we get a fleet that is serviceable for the long haul.

TC: That’s all I’ve got here. Thanks again for taking the time to chat.

You can watch a recap of the recent Virgin Galactic launch here.

Researchers match DeepMind’s AlphaFold2 protein folding power with faster, freely available model

DeepMind stunned the biology world late last year when its AlphaFold2 AI model predicted the structure of proteins (a common and very difficult problem) so accurately that many declared the decades-old problem “solved.” Now researchers claim to have leapfrogged DeepMind the way DeepMind leapfrogged the rest of the world, with RoseTTAFold, a system that does nearly the same thing at a fraction of the computational cost. (Oh, and it’s free to use.)

AlphaFold2 has been the talk of the industry since November, when it blew away the competition at CASP14, a virtual competition between algorithms built to predict the physical structure of a protein given the sequence of amino acids that makes it up. The model from DeepMind was so far ahead of the others, so highly and reliably accurate, that many in the field have talked (half-seriously and in good humor) about moving on to a new field.

But one aspect that seemed to satisfy no one was DeepMind’s plans for the system. It was not exhaustively and openly described, and some worried that the company (which is owned by Alphabet/Google) was planning on more or less keeping the secret sauce to themselves — which would be their prerogative but also somewhat against the ethos of mutual aid in the scientific world.

Update: DeepMind published more detailed methods in the journal Nature today. The code is available on GitHub. This does considerably lessen the aforementioned concern, but the advance described below is still highly relevant.

That concern seems to have been at least partly mooted by work from University of Washington researchers led by David Baker and Minkyung Baek, published in the latest issue of the journal Science. Baker, you may remember, recently won a Breakthrough Prize for his team’s work combating COVID-19 with engineered proteins.

The team’s new model, RoseTTAFold, makes predictions at similar accuracy levels using methods that Baker, responding to questions via email, candidly admitted were inspired by those used by AlphaFold2.

“The AlphaFold2 group presented several new high level concepts at the CASP14 meeting. Starting from these ideas, and with a lot of collective brainstorming with colleagues in the group, Minkyung has been able to make amazing progress in very little time,” he said. (“She is amazing!” he added.)

Examples of predicted protein structures and their ground truths. A score above 90 is considered extremely good.

Baker’s group more or less placed second at CASP14, no mean feat, but hearing DeepMind’s methods described even generally set them on a collision course. They developed a “three-track” neural network that simultaneously considered the amino acid sequence (one dimension), distances between residues (two dimensions), and coordinates in space (three dimensions). The implementation is beyond complex and far outside the scope of this article, but the result is a model that achieves almost the same accuracy levels — levels, it bears repeating, that were completely unprecedented less than a year ago.

What’s more, RoseTTAFold accomplishes this level of accuracy far more quickly — that is, using less computation power. As the paper puts it:

DeepMind reported using several GPUs for days to make individual predictions, whereas our predictions are made in a single pass through the network in the same manner that would be used for a server…the end-to-end version of RoseTTAFold requires ~10 min on an RTX2080 GPU to generate backbone coordinates for proteins with less than 400 residues.

Hear that? It’s the sound of thousands of microbiologists sighing in relief and discarding drafts of emails asking for supercomputer time. It may not be easy to lay one’s hands on a 2080 these days, but the point is any high-end desktop GPU can perform this task in minutes, instead of requiring a high-end cluster running for days.

The modest requirements make RoseTTAFold suitable for public hosting and distribution as well, something that might never have been in the cards for AlphaFold2.

“We have a public server that anyone can submit protein sequences to and have the structures predicted,” Baker said. “There have been over 4500 submissions since we put the server up a few weeks ago. We have also made the source code freely available.”

This may seem very niche, and it is, but protein folding has historically been one of the toughest problems in biology and one towards which countless hours of high-performance computing have been dedicated. You may recall [email protected], the popular distributed computing app that let people donate their computing cycles to attempting to predict protein structures. The kind of problem that might have taken a thousand computers days or weeks to do — essentially by brute-forcing solutions and checking for fit — now can be done in minutes on a single desktop.

The physical structure of proteins is of utmost importance in biology, as it is proteins that do the vast majority of tasks in our bodies, and proteins that must be modified, suppressed, enhanced, and so on for therapeutic reasons; first, however, they need to be understood, and until November that understanding could not be reliably achieved computationally. At CASP14 it was proven to be possible, and now it has been made widely available.

It is not, by a long shot, a “solution” to the problem of protein folding, though the sentiment has been expressed. Most proteins at rest in neutral conditions can now have their structure predicted, and that has huge repercussions in multiple domains, but proteins are seldom found “at rest in neutral conditions.” They twist and contort to grab or release other molecules, to block or slip through gates and other proteins, and generally to do everything they do. These interactions are far more numerous, complex, and difficult to predict, and neither AlphaFold2 nor RoseTTAFold can do so.

“There are many exciting chapters ahead… the story is just beginning,” said Baker.

If you’re curious about the science and the potential repercussions, consider reading this much more detailed and technical account of the methods and possible next steps written in the wake of AlphaFold2’s CASP14 performance.

Facebook is shook, asks for removal of FTC Chair Khan from antitrust cases against it

Facebook has joined Amazon in a show of alarm at the sudden rise of antitrust hawk Lina Khan to the position of FTC Chair by asking that she be recused from all decisions relating to the company. The argument, more or less identical to Amazon’s, is that before her appointment, Khan was too outspoken about her professional opinion that companies like these are in violation of antitrust rules.

In a letter filed with the FTC and obtained by the WSJ, which the agency could not provide and declined to comment on, Facebook explained that Khan’s last few years of academic publications and articles in other media amount to cause for recusal from decisions about the company. (I have asked Facebook for a copy of the petition and will update this post if I receive it.)

“Chair Khan has consistently made public statements not only accusing Facebook of conduct that merits disapproval but specifically expressing her belief that the conduct meets the elements of an antitrust offense. When a new commissioner has already drawn factual and legal conclusions and deemed the target a lawbreaker, due process requires that individual to recuse herself,” reads the petition.

Neither the FTC nor Khan in any other capacity have responded to the recusal requests from Facebook and Amazon. She did note in her nomination proceeding that recusal requests like these do happen, and are resolved on a case-by-case basis (unlike automatic recusals for things like financial or personal interest). Perhaps even now she is meeting with the ethics experts at the agency.

Khan has, however, certainly made her policy positions known in numerous articles and papers, many of which have argued that antitrust regulators have been highly conservative in their interpretation and deployment of their legal powers, and equally permissive in their oversight of the current crop of enormous tech companies. Things like acquiring competitors, artificially lowering prices to pressure a market, or misrepresenting the collection and use of customer data have gone either unchallenged or minimally punished.

In particular she acted as counsel for the House’s Investigation of Competition in Digital Markets, an antitrust report issued last fall. Amazon and Facebook lean on cases from 1966 and 1970 where an FTC Commissioner was recused for “prejudgment” of a case during a Congressional investigation in which he participated. It’s a promising hook to hang a case on to be sure, but the circumstances are by no means equivalent. I’m not a lawyer, but it seems to me that no case or even specific allegations have been prejudged, only the general idea that Facebook, Apple, Google, and Amazon all either have monopolies or otherwise possess market power. (They didn’t care much when the report was issued.)

The main finding of the House report, in fact, was arguably that there could be no legal case because existing laws and regulations are insufficient. Certainly Khan has shouted this from the rooftops for some years now — but the conclusion is a legislative matter, not an FTC one. It would be mighty difficult for Khan to have prejudged an antitrust case predicated on laws that haven’t yet been written.

Khan’s FTC has suffered an early setback on her watch though not of her making in the dismissal of some complaints in the agency’s current antitrust case against Facebook. It was for lack of evidence that the company exerts monopoly control over social media that the judge told the FTC to come back and try again. Perhaps Khan intends to remedy that with a supplemented filing, or perhaps she will take the loss and muster her forces for another go in a year or two — but either way it is probably best to resolve the question of her alleged “prejudgment” before that decision is announced. (The FTC declined to speculate as to whether the recusal request would affect the current proceedings.)

But the agency also has explicit backing from the White House in the form of President Biden’s request that it prioritize “dominant internet platforms, with particular attention to the acquisition of nascent competitors, serial mergers, the accumulation of data, competition by ‘free’ products, and the effect on user privacy.” So Khan probably isn’t feeling the sting of the aforementioned legal challenge.

The petitions filed by Amazon and Facebook have near-zero risk for the companies and an outside chance at provoking a recusal, so it makes sense strategically to file them. They also provide breadcrumbs later for their inevitable objections to the FTC’s (under Khan, equally inevitable) allegations of monopolistic practices. The legal repercussions are hard to predict but it is usually better to have a complaint on the table already rather than bring it out late in the process.

Given Chair Khan’s position that the FTC itself needs to be overhauled and empowered in order to bring actions like this against companies like Facebook, it seems clear that all these are merely the opening gambits in a long, long game.

Cutting out carbon emitters with bioengineering at XTC Global Finals on July 22

Bioengineering may soon provide compelling, low-carbon alternatives in industries where even the best methods produce significant emissions. Utilizing natural and engineered biological process has led to low-carbon textiles from Algiknit, cell-cultured premium meats from Orbillion, and fuels captured from waste emissions via LanzaTech — and leaders from those companies will be joining us on stage for the Extreme Tech Challenge Global Finals on July 22.

We’re co-hosting the event, with panels like this one all day and a pitch-off that will feature a number of innovative startups with a sustainability angle.

I’ll be moderating a panel on using bioengineering to create change directly in industries with large carbon footprints: textiles, meat production, and manufacturing.

Algiknit is a startup that is sourcing raw material for fabric from kelp, which is an eco-friendly alternative to textile crop monocultures and artificial materials like acrylic. CEO Aaron Nessa will speak to the challenge of breaking into this established industry and overcoming preconceived notions of what an algae-derived fabric might be like (spoiler: it’s like any other fabric).

Orbillion Bio is one of the new crop of alternative protein companies offering cell-cultured meats (just don’t call them “lab” or “vat” grown) to offset the incredibly wasteful livestock industry. But it’s more than just growing a steak — there are regulatory and market barriers aplenty that CEO Patricia Bubner can speak to as well as the technical challenge.

LanzaTech works with factories to capture emissions as they’re emitted, collecting the useful particles that would otherwise clutter the atmosphere and repurposing them in the form of premium fuels. This is a delicate and complex process that needs to be a partnership, not just a retrofitting operation, so CEO Jennifer Holmgren will speak to their approach convincing the industry to work with them at the ground floor.

It should be a very interesting conversation, so tune in on July 22 to hear these and other industry leaders focused on sustainability discuss how innovation at the startup level can contribute to the fight against climate change. Plus it’s free!

Virgin Galactic and Richard Branson celebrate launch of first passengers into space

Virgin Galactic has successfully taken its first passengers to space, including its billionaire founder Richard Branson. The event, at Spaceport America in New Mexico, was a field day for press and employees, complete with an early-morning Khalid set and hero walk by Branson and the crew.

“Just imagine a world where people of all ages, all backgrounds, from anywhere, of any gender, of any ethnicity, have equal access to space,” Branson said on returning. “Welcome to the dawn of a new space age!”

The remark is a bit premature, of course — that world is still some distance off, but it’s true that this flight marks a historic moment in the nascent space tourism industry. At present, leisurenauts are still an elite class, but the events of the day suggest we’re closer than ever to seeing that change.

After an incredibly early start to the day (shuttles to the Spaceport left at 2:45 AM from nearby Las Cruces), the festivities began in true space launch style with a delay. A thunderstorm overnight prevented the team from rolling out the spacecraft, which believe it or not can’t get wet. At the speeds and temperatures involved nothing can be left to chance — like ice forming from water in or on the chassis.

Press set up before dawn at Spaceport America.

Image Credits: Devin Coldewey / TechCrunch

Soon the sun rose and crowds arrived: VIPs, employees, a bunch of local students, and Branson’s own guest list (reportedly numbering around 150). Elon Musk showed up as well, presumably to congratulate his fellow spaceman personally, billionaire to billionaire.

At 8:30 local time the engines started on VMS Eve, the “mothership” carrying VSS Unity, the rocket-powered spaceplane that Branson, along with Virgin Galactic’s Beth Moses (her second flight), Sirisha Bandla, and Colin Bennett, would ride to the edge of space.

VMS Eve takes off. Image Credits: Virgin Galactic

Eve was wheels up at 8:40, commencing a wait on the ground while it climbed to about 36,000 feet. Unity detached and began its rocket-powered climb at about 9:24, reaching Mach 3 and after two minutes reached its peak altitude of about 282,000 feet — about 53 miles, as planned.

The crew and passengers enjoyed a minute or two of microgravity, which they seem to have employed gainfully:

Image Credits: Virgin Galactic

A planned mid-air speech by Branson proved impossible as the signal cut in and out, but the craft itself proved more reliable, touching down at 9:38.

In a celebratory stage appearance (following a brief Khalid concert) Branson expanded on the ideas cut short in transmission, beginning with: “It’s hot, I’m sorry,” but quickly moving on to more inspiring words. “I have dreamt about this moment since I was a child, but nothing could have prepared me for the view of Earth from space. We are at the vanguard of a new space age.”

At a press conference following shortly after, Branson fielded questions from elementary schoolers, and the crew described the view from space and whether they saw any planets. (No, just an alien that the pilot shook off during descent, Branson said. At least one kid I saw believed him.)

A long road to space

Virgin Galactic Pilots on their way to the Virgin Galactic Spaceflight System

It’s a triumph long in the making for Virgin Galactic and Branson. The company was ahead of the curve in its space tourism ambitions, but in 2014 a test flight ended in a horrific crash and the death of one of the pilots.

Virgin’s engineers and leaders worked through it, however, and built a stronger, better spacecraft which was christened Unity by Stephen Hawking, who was then still living — and, not surprisingly, hoping to hitch a ride some day.

Pilots flew test flight after test flight over the years, slowly ratcheting up the power and finally, in 2018, touching the edge of space. On that note there is some slight controversy in that the exact altitude where the atmosphere gives way to space isn’t completely agreed upon. Some authorities place the Kármán line, as the imaginary boundary is called, at 100 kilometers above sea level, others at 50 miles, or about 80 kilometers.

Unity 22 spreads its “feathers” during descent. Image Credits: Virgin Galactic

Virgin uses the lower estimate, while its arch rival, Jeff Bezos’s Blue Origin, uses the higher. This led Bezos to throw shade on Virgin’s flights, saying he didn’t want his customers to have an “asterisk” on their trip to space. When I asked about this before, a Virgin representative said they use the same standard that NASA and the U.S. Air Force does: pilots are given their “astronaut wings” if they pass the 50-mile mark.

Kármán quibbles aside, the race to send passengers to space has been heating up lately, and Bezos recently announced that he would be flying aboard the first crewed launch of Blue Origin’s New Shepard rocket on July 22 — with his brother, a mystery passenger who has paid $28M for the privilege, and Wally Funk, among the first women trained to be astronauts in 1961 but who never made it to space.

But Branson rained on his parade by announcing shortly afterward that he would fly aboard Virgin’s first passenger launch to space (crew and pilots have been up several times) about a week earlier.

While Branson has good-naturedly denied any competition between himself and Bezos (“We wish jeff the absolute best,” he said, adding that Bezos sent over a message of goodwill before the flight) it’s hard to believe that’s completely true. Though neither man has anything to prove at this point, there must surely be some satisfaction in Branson’s not merely going to space (a lifelong dream, as he tells it) but doing so before his upstart rival. However much he denies it, the narrative is too tempting to quash completely.

The direction forward for Virgin Galactic now is, clearly, towards paying customers, of which there are plenty lined up. Of course, they all have a quarter of a million dollars to spare, but you might not, and for you Branson has a special offer. They’ve partnered with Omaze, and donations to the chosen charity will enter you into a raffle of sorts, with the winner receiving two tickets on an upcoming Virgin Galactic flight. “And with my Willy Wonka hat on, a guided tour of Spaceport America, given by yours truly,” Branson added.

Branson expressed hope that this would become an ongoing thing as long as donations continue, so perhaps this is the answer to the question of how they hope to, as he so frequently promises, make space available to everyone.

You can watch the whole day unfold as it happened in Virgin Galactic’s archived livestream below:

Watch live as Virgin Galactic’s first passenger flight takes off with Richard Branson on board

Virgin Galactic is set to launch its first passengers to space tomorrow morning, and you can watch the whole thing right here. The launch is scheduled for 6 AM Pacific, with streaming festivities (including commentary by Stephen Colbert) starting on the hour.

This launch is the 22nd for VSS Unity, Virgin Galactic’s first spacecraft to leave the atmosphere. As before, Unity will leave the Spaceport attached to the belly of VMS Eve, which will take it up out of the thickest part of the atmosphere.

Unity will drop off and ignite its rocket engine, reaching speeds approaching Mach 3 until it reaches the 80 kilometer mark, the lowest altitude considered to be in space. When the engines shut off, the pilots and passengers will enjoy a short period of weightlessness and of course stunning views.

The whole thing will be livestreamed, from the ground and, connectivity permitting, from the craft itself. When they return safely there will be a triumphant press conference and, remarkably, live music from Khalid.

I’ll be there on the ground, but you can see what I see by tuning in to the official stream below:

Watch live as Virgin Galactic’s first passenger flight takes off with Richard Branson on board

Virgin Galactic is set to launch its first passengers to space tomorrow morning, and you can watch the whole thing right here. The launch is scheduled for 6 AM Pacific, with streaming festivities (including commentary by Stephen Colbert) starting on the hour.

This launch is the 22nd for VSS Unity, Virgin Galactic’s first spacecraft to leave the atmosphere. As before, Unity will leave the Spaceport attached to the belly of VMS Eve, which will take it up out of the thickest part of the atmosphere.

Unity will drop off and ignite its rocket engine, reaching speeds approaching Mach 3 until it reaches the 80 kilometer mark, the lowest altitude considered to be in space. When the engines shut off, the pilots and passengers will enjoy a short period of weightlessness and of course stunning views.

The whole thing will be livestreamed, from the ground and, connectivity permitting, from the craft itself. When they return safely there will be a triumphant press conference and, remarkably, live music from Khalid.

I’ll be there on the ground, but you can see what I see by tuning in to the official stream below: